everyone posted about the revolutionary H3 Max by @fal, which made AI livestreams possible.
@levelsio even hooked it to an actual stream. then the fal team followed.
so i spent the weekend building the missing link: launching your own channels.
📺 → https://t.co/yo0yjyrCak
@OriSilver Can’t wait to use this in maxfusion, I currently have separate processes for this that my Claude Code has to manually manage alongside maxfusion.
Doing this natively will be crazy useful 👀
GPT-Live-1 is now available in the API.
Bring ChatGPT’s natural back-and-forth to your app, with voice agents that listen while they speak and work with the models and harness you choose.
i don't think tech people understand how little non-tech people care about tech
the average person does not have a basic understanding of how a computer works
the idea of everyone just prompting their own software is only for a small subset of people
Ad Delayer from @bria_ai_ is now live on fal
- Recovers a layered, editable composition from a finished flat ad
- Headlines come back as SVG text, so localizing a variant is a structural edit, not a re-creation
- Layers arrive typed and ordered
@mfishbein@canva Exactly.
But also for any ad, even yours.
Let’s say you generate an ad from Nano Banana Pro, you can still use it to make the static .png file a dynamic editable file
🌍 What if an AI-generated world didn’t end after a few seconds—but kept running for hours, responded to every move, let you attack, cast spells, shoot, or summon storms, and continued evolving through AI agents?
That’s LingBot-World 2.0: our open-source real-time interactive world model, scaling to 720p/60fps in our full setup.
Today, we’re open-sourcing 3 more models:
🔹 LingBot-World 2.0 Small (1.3B)
🔹 LingBot-World 2.0 Bidirectional
🔹 LingBot-World 2.0 Causal Pretrain
And now, with our 1.3B Small Model, that world can run in real time on a single consumer GPU. 🧵
You now get the same distribution power once reserved for lobbying agencies..
Introducing TokPortal: build a fleet of human-managed, geo-targeted social accounts and distribute your content worldwide.
now starts a short period of time when creative people who are extremely good at prompting are so much more productive at making games, shorts, movies, and other content, that they can compete starting from zero and make new empires. maybe a year until people catch up
📺 https://t.co/1FE90eHWbM's day 8
┣ 293 clips aired · 262 from viewers · $37.38 burned
┣ 267 prompts reached air, 51 turned away at the gate
┣ interact rate 18.6%
┣ channel claims 10
┗ clip of the day ↓
"Channel Blueprint: BrandLift360
Channel Identity & Framing
* Handle / Title: @ BrandLift360
* Headline: Answer Engine Optimization, Autonomous GTM, and the AI Workforce."
somebody pasted a whole channel blueprint into the prompt box, all 200 characters it would take, and the channel built the blueprint. he found the real claim form five minutes later and filled it in properly.
the site was also down for 105 minutes while row level security went on the database. nobody was on it for any of that, which is luck rather than planning.
whatever you type is what airs next, and if it's good it's in tomorrow's post :)
H3 Max Director 1.1 is now available.
New features for longer, more controllable video generation:
- Start with any frame and insert keyframes throughout a generation
- Introduce audio at any point, with native audio conditioning
- Generate natively in 1080p
- Run continuous generations for up to 15 minutes
Director adapts to new visual and audio inputs in real time, giving you precise control without breaking continuity.
@eptwts Exactly, that’s what https://t.co/yo0yjyrCak is trying to build
Basically the YouTube of interactive AI media, letting anyone launch their channels with no technical knowledge
We’re extending our 75% off promotional pricing across all currently discounted H3 Max and H3 Max Turbo endpoints through September 15.
For teams running H3 Max at scale, contact [email protected] to discuss permanent volume discounts.